Do I have a duty to tell clients I use AI?

There is no blanket duty — ABA Formal Opinion 512 and the state bars agree on that. Disclosure is required in specific situations: when the client asks, when the engagement agreement or outside counsel guidelines require it, when informed consent is needed under Rule 1.6 because client information will be input into a self-learning tool, when AI use is relevant to the basis or reasonableness of the fee, and when AI output will influence a significant decision in the representation.

When do I need client consent before using AI on client information?

Under Rule 1.6, if you input information relating to the representation into a self-learning GAI tool, the client's informed consent is required in advance — and boilerplate engagement-letter language is not enough; the consent must explain the tool, the risks, what information is shared, and the benefits. If the tool is not self-learning or you are not inputting client information, consent is generally not required.

Do state bars agree on when disclosure is required?

No — they diverge meaningfully. Utah and West Virginia require consent before using generative AI tools; Pennsylvania requires informing clients how tools are used and disclosing AI expenses; Kentucky exempts 'routine use' unless clients are charged or a court rule requires it; D.C. ties disclosure to billing for AI costs; California and Florida weigh the facts and risks case by case. The safe approach is the one New York suggests: a plain-language statement in the engagement letter.

What is the safest disclosure practice?

Disclose in writing at the outset. A plain-language paragraph in the engagement letter describing the categories of AI tools you may use, what happens to client information, and how AI affects billing — plus an invitation to ask for detail — satisfies the disclosure rules in every state, avoids the boilerplate-consent trap under Rule 1.6, and protects you in the jurisdictions with the strictest requirements. Clients hire lawyers, not computers, and most appreciate knowing how the work is done.

Disclosure Duties: Do You Have to Tell Clients You Use AI?

Q: Do courts require disclosure of AI use?

Increasingly, yes. The first federal judge required AI disclosure in May 2023; by 2026, more than 300 federal judges had standing orders or local rules requiring some form of AI disclosure or certification, and the Ropes & Gray tracker recorded 681 court orders and rules by mid-2026 — up from 36 across 13 states in mid-2024. These rules vary widely by judge, which is why litigators must check the specific court's requirements before filing.

Every lawyer using AI has faced the question from a partner, a client, or their own conscience: do I have to tell them? The honest answer is more useful than the comfortable one. There is no blanket ethical duty to disclose every use of AI — the ABA and the state bars agree on that. But there are five situations where disclosure is required, a growing list of court rules that impose their own requirements, and a state-by-state landscape where the bars genuinely disagree. This article maps all three, so you know what you must disclose, what you should disclose, and where the answer depends on where you practice.

It builds on the framework in our guide to ABA Formal Opinion 512 at your desk — this article takes the communication thread and follows it further.

The starting point: no blanket duty

The consensus across every authority is that AI use, by itself, does not need to be disclosed. ABA Formal Opinion 512 is explicit that disclosure questions are "factually specific" and that there is no one-size-fits-all answer. The state bars that have addressed the question — California, Florida, Kentucky, Texas, New York — all start from the same place: using AI is not inherently a material fact requiring disclosure, any more than using a particular research service or a word processor is. The duty to communicate under Rule 1.4(a)(2) — to "reasonably consult with the client about the means by which the client's objectives are to be accomplished" — is triggered by circumstances, not by the mere fact of the tool.

The five situations where disclosure is required

Opinion 512 lists the situations where disclosure is not optional. If any of these apply, you must communicate with the client about AI use:

The client asks. If a client asks whether you used AI, you must answer truthfully and completely — this is Rule 1.4 at its most basic, and there is no discretion here.
The engagement agreement or outside counsel guidelines require it. If your contract or the client's guidelines impose disclosure, the contract controls. This is increasingly common as sophisticated clients add AI-use provisions to their outside counsel guidelines.
Informed consent is required under Rule 1.6. When you input information relating to the representation into a self-learning GAI tool, the client's informed consent is required in advance — and that consent process is itself a disclosure.
AI use is relevant to the basis or reasonableness of the fee. If you bill for AI-assisted work or pass through AI-related costs, the client is entitled to understand what they are paying for.
The tool's output will influence a significant decision in the representation. If AI output shapes strategy, settlement, or another significant choice, the client should know the basis for the decision — the same logic that governs any significant development.

Beyond these five, the opinion says the analysis turns on factors like the client's needs and expectations, the scope of the representation, the sensitivity of the information, the tool's importance to the task, and — in a phrase worth keeping — "the extent to which knowledge of the lawyer's use of the tool would affect the client's evaluation of or confidence in the lawyer's work." Clients hire lawyers, not computers; if knowing how the work was done would change the client's assessment, that is a strong signal disclosure is warranted.

Where the state bars diverge

The national consensus ends at "no blanket duty." On the specifics, the states genuinely differ, and where you practice changes the answer:

Utah — the strictest early entrant: seek and obtain client consent before using generative AI tools such as ChatGPT (Utah State Bar, 2023).
West Virginia — consult with clients before using AI tools and obtain consent in writing (Legal Ethics Opinion 24-01, 2024).
Pennsylvania — must inform clients of AI use, explain how tools are used and their impact on case objectives, and disclose AI-related expenses (Joint Formal Opinion 2024-200, 2024).
Kentucky — the most permissive: "routine use" of AI need not be communicated, unless the client is being charged for AI-related costs or a court rule requires it (KBA E-457, 2024).
District of Columbia — ties disclosure to billing: a duty to communicate AI use if the lawyer intends to bill the client for AI-related out-of-pocket costs (Ethics Opinion 388, 2024).
California — no strict requirement, but lawyers should weigh the duty "based on the facts and circumstances, including the novelty of the technology, risks associated with generative AI use, scope of the representation, and sophistication of the client" (Practical Guidance, Nov. 2023).
Florida — consent obligations depend on risk and client expectations; consent is required where client confidential information will be input into the tool (Ethics Opinion 24-1, Jan. 2024).
New York — recommends including in the retainer agreement a statement that AI tools may be utilized, with the client's acknowledgement (NYSBA Task Force report, April 2024).

The practical consequence: a firm practicing in one state, or admitted in several, cannot assume a single national answer. The exposure is real — an attorney admitted in multiple states must comply with each state's rules, and a practice that satisfies the permissive reading in one jurisdiction may fall short in another.

The courts are building their own regime

While the ethics framework was settling, the courts built a parallel disclosure system. The first federal judge required AI disclosure in May 2023. The pace since then is the story: by mid-2024, tracking by Ropes & Gray counted at least 36 standing orders from state and federal judges across 13 states; by 2026, more than 300 federal judges had standing orders or local rules requiring some form of AI disclosure or certification, and the Ropes & Gray tracker recorded 681 court orders and rules by mid-2026.

The requirements vary wildly by judge — some require disclosure of any AI use in drafting, some require certification that AI-generated citations were verified by a human, some require disclosure of which tool was used and confirmation that no confidential information was exposed. Judge Brantley Starr of the Northern District of Texas was the first to require a certificate attesting that AI-generated content was verified by a human. The through-line is that none of these rules cares what your state ethics opinion says: if the judge in your case requires disclosure, the judge's rule controls your filing. Litigators must check the specific court's requirements before filing — a checklist we covered in verifying an AI's case citation.

The temporary-lawyer analogy

The bar's own history offers a useful lens. Before AI, the same disclosure question arose about temporary and contract lawyers. Virginia saw "little purpose" in disclosing support services that are "truly tangential, clerical, or administrative in nature." California requires disclosure when the temporary lawyer's work is a "significant development" in the case. New York weighs whether client confidences will be shared, the lawyer's degree of involvement, and the significance of the work. Florida asks how important the use is in the eyes of the client. The ABA suggested these opinions as the analogy for AI — and the pattern holds: disclose when the tool's role is significant, protects confidences, or affects the client's understanding of the work; no disclosure needed for the clerical and administrative.

The safe practice: disclose in writing, at the outset

Here is the thing about the divergent landscape: the safe answer is the same everywhere. Add a plain-language paragraph to your engagement letter describing the categories of AI tools you may use, what happens to client information, and how AI affects billing — and invite the client to ask for detail. This single practice:

Satisfies every state's disclosure rule — the strictest states (Utah, West Virginia, Pennsylvania) are covered by the upfront disclosure, and the permissive states are not harmed by it.
Avoids the Rule 1.6 boilerplate trap — because the disclosure is specific about tools, data, and risks, it can double as the informed-consent language when client information enters a self-learning tool.
Meets the fee-transparency requirement — clients know AI is part of the workflow and what, if anything, it costs them.
Builds client trust — most clients respond well to transparency about how their work is done; surprise is the thing that erodes confidence.

One caution: the paragraph must be specific enough to be real. A generic "we may use AI tools" sentence is exactly the kind of boilerplate Opinion 512 warns is insufficient when informed consent is required. Name the categories, state the data practices, and offer the detail.

What Lawyer Assistant does about this

Lawyer Assistant's design makes the disclosure conversation easy to have honestly. Because everything runs locally on your machine, the data-practices answer is short and true: no vendor access, no training on your data, no cloud retention. That is the kind of specific, verifiable statement an engagement-letter paragraph — and an informed-consent conversation — can actually make. The tool does not change your disclosure duties; it gives you a clean, truthful thing to disclose. The duties themselves, as always, belong to you.

The bottom line

Do you have to tell clients you use AI? Sometimes. The honest map: no blanket duty, but five situations where disclosure is mandatory; a state landscape that ranges from "consent in writing before use" (West Virginia) to "routine use needs no notice" (Kentucky); and a court system that by 2026 had more than 300 federal judges imposing their own disclosure requirements. The practice that survives all three regimes is the same: disclose in writing at the outset, be specific, and tell the truth when asked. That is not the cautious answer — it is the only answer that works in every jurisdiction at once.

Sources & further reading

ABA Standing Committee on Ethics and Professional Responsibility, Formal Opinion 512, "Generative Artificial Intelligence Tools" (July 29, 2024) (PDF) — the five disclosure-trigger situations and the "factually specific" framework.
State bar AI guidance: Utah State Bar (2023); West Virginia Legal Ethics Opinion 24-01 (2024); Pennsylvania/Philadelphia Joint Formal Opinion 2024-200 (2024); Kentucky KBA E-457 (2024); D.C. Bar Ethics Opinion 388 (2024); California Practical Guidance on Generative AI (Nov. 2023); Florida Bar Ethics Opinion 24-1 (Jan. 2024); NYSBA Task Force on AI report (April 2024) — summarized in Esquire Deposition Solutions, "Litigators Weigh Need to Disclose AI Use to Clients".
Ropes & Gray LLP, AI Court Order Tracker — 36 standing orders across 13 states as of mid-2024; 681 court orders and rules by mid-2026; more than 300 federal judges with AI disclosure requirements by 2026.
Esquire Deposition Solutions, "Litigators Weigh Need to Disclose AI Use to Clients" — the state-by-state summary and the temporary-lawyer analogy.

This article is general information about technology and professional practice. It is not legal advice for any specific matter, and rules vary by jurisdiction — verify against the authority applicable to your matter.